Automated Synthetic Scene Generation
Abstract
First principles, physics-based models help organizations developing new remote sensing instruments anticipate sensor performance by enabling the ability to create synthetic imagery for proposed sensor before a sensor is built. One of the largest challenges in modeling realistic synthetic imagery, however, is generating the spectrally attributed, three-dimensional scenes on which the models are based in a timely and affordable fashion. Additionally, manual and semi-automated approaches to synthetic scene construction which rely on spectral libraries may not adequately capture the spectral variability of real-world sites especially when the libraries consist of measurements made in other locations or in a lab. This dissertation presents a method to fully automate the generation of synthetic scenes when coincident lidar, Hyperspectral Imagery (HSI), and high-resolution imagery of a real-world site are available. The method, called the Lidar/HSI Direct (LHD) method, greatly reduces the time and manpower needed to generate a synthetic scene while also matching the modeled scene as closely as possible to a real-world site both spatially and spectrally. Furthermore, the LHD method enables the generation of synthetic scenes over sites in which ground access is not available providing the potential for improved military mission planning and increased ability to fuse information from multiple modalities and look angles. The LHD method quickly and accurately generates three-dimensional scenes providing the community with a tool to expand the library of synthetic scenes and therefore expand the potential applications of physics-based synthetic imagery modeling.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2014
- Accession Number
- ADA608884
Entities
People
- Ryan N. Givens
Organizations
- Air Force Institute of Technology